Objectives: Prednisolone therapy increases the risk of infections in patients with severe alcoholic hepatitis (SAH). We evaluated whether the use of the Lille Model at day 4 (LM4) is useful to predict response to prednisolone compared with the classic day 7 (LM7) in order to limit a futile exposure to corticosteroids.
Methods: We performed a retrospective analysis of a large multinational cohort of patients with SAH with Maddrey's discriminant function (DF) ≥32. Response to corticosteroids was assessed with LM4 and LM7, according to the validated cutoff value (CUV>0.45). Receiver operating characteristics (ROC) curves were constructed to determine the optimal CUV for LM4 and to compare accuracy between LM4, LM7, MELD (Model for End-Stage Liver Disease), and ABIC (age, bilirubin, international normalized ratio, and creatinine). Logistic regression models were constructed to predict 28- and 90-day mortality. Cox regression analysis was performed to assess long-term survival.
Results: A total of 163 (62.7%) out of 260 patients received corticosteroids. The median DF for the patients treated with corticosteroids was 64.1 (47.9-81.3). Overall 90-day mortality was 35.9%. The median LM4 and LM7 for the patients who received treatment was 0.39 (0.19-0.83) and 0.36 (0.13-0.77). LM4 was a strong independent predictor of 28-day mortality (OR 25.4, (95% confidence interval (CI) 5.1-126.8), P<0.001). By using LM4 with a CUV>0.45, 28- and 90-day survival was significantly higher for responders (90% and 76%) than non-responders (66% and 40%), P<0.001. Importantly, the area under the ROC curve for predicting mortality for LM4 was similar than the classic LM7 (0.77 vs. 0.75, respectively: P=0.558).
Conclusions: LM4 is as accurate as LM7 in predicting response to corticosteroids, as well as 28- and 90-day mortality. Assessing the efficacy of prednisolone at an earlier time point can avoid a more prolonged futile use of this therapy.